CN108062757A - It is a kind of to utilize the method for improving Intuitionistic Fuzzy Clustering algorithm extraction infrared target - Google Patents
It is a kind of to utilize the method for improving Intuitionistic Fuzzy Clustering algorithm extraction infrared target Download PDFInfo
- Publication number
- CN108062757A CN108062757A CN201810010258.8A CN201810010258A CN108062757A CN 108062757 A CN108062757 A CN 108062757A CN 201810010258 A CN201810010258 A CN 201810010258A CN 108062757 A CN108062757 A CN 108062757A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- degree
- msup
- mover
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
- Other Investigation Or Analysis Of Materials By Electrical Means (AREA)
- Secondary Cells (AREA)
- Medicines Containing Antibodies Or Antigens For Use As Internal Diagnostic Agents (AREA)
Abstract
The present invention proposes a kind of method for extracting infrared target using Intuitionistic Fuzzy Clustering algorithm is improved, and approximate region of the infrared target in infrared image is determined first with conspicuousness algorithm.Improved Intuitionistic Fuzzy Clustering algorithm is recycled to carry out cluster segmentation to the region.Finally, nontarget area is rejected by simply post-processing.Local region information and non local symmetry information are taken into full account in innovatory algorithm, therefore segmentation result has obtained apparent improvement.Specially:Step 1:Determine target approximate region.The notable figure of infrared image is obtained using conspicuousness algorithm, by obtaining target approximate location into row threshold division to notable figure.Step 3:The mirror symmetry detection method based on registration proposed using Marcelo Xi Keen et al. completes target symmetry axis detecting step four:Cluster segmentation is carried out to infrared image using improved Intuitionistic Fuzzy Clustering algorithm.Step 5:To the image segmentation result that step 4 obtains into subsequent processing, nontarget area is rejected.
Description
【Technical field】
The method for improving Intuitionistic Fuzzy Clustering algorithm extraction infrared target, Fuzzy clustering techniques are utilized the present invention relates to a kind of
It is had a wide range of applications with image Segmentation Technology in image application field, is under the jurisdiction of digital image processing field.
【Background technology】
Image segmentation refers to be drawn original image according to certain or some features (gray scale, texture, histogram etc.) of image
It is divided into the region of several non-overlapping copies.Similarity is higher inside the same area, and it is apparent poor to have between different zones
It is different.Image partition method is broadly divided into following a few classes according to principle difference:Dividing method based on threshold value, point based on region
Segmentation method, the dividing method based on edge and dividing method based on specific theory etc..
Symmetry is a kind of common characteristic of object.Have in many of life object on some direction to a certain degree
Symmetry characteristic, such as pedestrian, automobile, naval vessel, aircraft, leaf etc..During image is split, on symmetrical axisymmetric
Pixel often there is larger possibility to belong to same category.Therefore, symmetric information is introduced into image segmentation and tended to
Enough improve the accuracy of image segmentation.
Infrared image can reflect the temperature difference of target and background in image, since it can all weather operations and can
The advantages of overcoming dysopia and detecting target, infrared image obtain in fields such as military affairs, industry, automobile assistant driving, medicine
To being widely applied, great concern is also resulted in image processing field.But infrared image often lacks preferable comparison
Degree and resolution ratio, soft edge, transitional stronger, boundary unobvious between target and background.Infrared image lacks texture
Information, and target is often for highlight regions, this causes the shape feature of target and symmetry characteristic seems and becomes apparent.Therefore, will
Symmetric information can improve it for infrared Image Segmentation and split accuracy.
In infrared image, target is often highlight regions, and therefore, the pixel for being under the jurisdiction of target shows significantly to gather
Class feature.Therefore, clustering algorithm has some superiority for infrared Image Segmentation.Traditional clustering algorithm belongs to rigid cluster, as
Vegetarian refreshments only belongs to Mr. Yu's class and is not belonging to two kinds of situations.And in fuzzy clustering method, data point is not drawing for hardness
It is a kind of to assign to certain, but multiclass is belonged to different degree.Therefore, it is substantially and uncertain high for transition in infrared image
Feature is split with certain advantage infrared image using the method for fuzzy clustering.
Fuzzy clustering algorithm has been obtained for being widely applied in image segmentation field.The classic algorithm of most common of which is
FCM Algorithms.Fuzzy C-mean algorithm (FCM) algorithm is (referring to document:J.C. Dunne one kind and the relevant mould of ISODATA algorithms
Paste image processing method and its applied to the compact easily separated cluster cybernetics journals of detection, 19733 (3):32-57.
(J.C.Dunn.A FuzzyRelative of the ISODATA Process and Its Use in Detecting
Compact Well-Separated Clusters[J].Journal of Cybernetics,1973,3(3):32-57.))
It is proposed at first by Dunne J.C. and is promoted through Bezdek, be a kind of Data Clustering Algorithm based on objective function optimization.But it passes
Do not consider spatial information and geological information in system FCM Algorithms, therefore to noise-sensitive, it cannot for noisy image
Enough obtain good segmentation result.For this problem, researcher proposes many improvement fuzzy clusterings for introducing spatial information
Algorithm.Such as Ahmed et al. considers the FCM_S algorithms of neighborhood information proposition (referring to document by introducing regular terms:It is solemn
Han Mode N Ahmeds, Sa Maiheyamani, interior text Mohammed, founder of Islam et al. one kind are estimated to change for bias-field
It is applied to .21 volumes of .193-199,2002. of MRI image segmentation .IEEE Medical Imagings into FCM Algorithms and its
(M.Ahmed,S.Yamany,N.Mohamed,A.Farag,and T.Moriarty,“Amodified fuzzy C-means
algorithm for bias field estimation and segmentation of MRI data,”IEEE
Trans.Med.Imag.,vol.21,pp.193–199,2002.));On the basis of FCM_S, Chen Songcan and road good general are equal
Value filtering and medium filtering thought, which are introduced into algorithm, further provides FCM_S1 and FCM_S2 algorithms, improves traditional fuzzy
The noiseproof feature of C averages is (referring to document:A kind of combining space information based on kernel function of stabilizations of Chen Songcan, Zhang Daoqiang obscures
C average image segmentation algorithm .IEEE system control process transactions .34 volumes of .1907-1916,2004. (S.Chen and
D.Zhang,“Robust image segmentation using FCM with spatial constraints based
on new kernel-induced distance measure,”IEEE Trans.Syst.,Man,Cybern.,vol.34,
pp.1907–1916,2004.));This safe Marcelo Rios and Wa Xilisi propose FLICM algorithms, which introduces new neighborhood
The information factor, while problem of parameter selection is avoided, make algorithm robustness more preferable.(referring to document:Konrad Staheli it is difficult to understand this, watt Seeley
At a kind of improvement FCM Algorithms American Institute of Electrical and Electronics Engineers images of stabilizations of combination local spatial information of this
Manage transactions in .19,1328-1337,2010 years Mays (S.Krinidis and V.Chatzis, A robust fuzzy local
information c-means clustering algorithm,IEEE Trans.Imag.Process.,vol.19,
no.5,pp.1328–1337,May 2010.));In addition, area information is introduced FCM algorithms to change by Lowe, Dewey K. K. et al.
The segmentation result of kind FCM algorithms is (referring to document:Lowe, Dewey K. K., Zhang Yun, king like a kind of improvement moulds of calmodulin binding domain CaM spatial information of people
Paste clustering algorithm for image split American Institute of Electrical and Electronics Engineers image procossing transactions, 3990-4000,2015 11
Moon (Guoying Liu, Yun Zhang, and Aimin Wang, Incorporating Adaptive Local
Information Into Fuzzy Clustering for Image Segmentation,IEEE
Trans.Imag.Process.,vol.24,no.11,pp.3990-4000,Novermber 2015));Xu Ze water et al. will be straight
Feel that fuzzy set theory is attached in FCM algorithms and propose intuitionistic fuzzy C averages, achieve good effect (referring to document:Xu Ze
Water, Wu person of outstanding talent intuitionistic fuzzy C mean algorithm system engineerings and electronic technology, volume 21,580-590, the of in August, 2010 (Zeshui
Xu,Junjie Wu.Intuitionistic fuzzy C-means clustering algorithms,
J.Syst.Eng.Electron,vol.21,pp.580-590,Aug 2010));Gu Jing et al. is by sparse from the same mould of representation theory
Paste C averages combine the innovatory algorithm proposed and achieve preferable segmentation result (referring to document:Gu Jing, burnt Lee into etc., based on sparse
From the improvement fuzzy C-mean algorithm American Institute of Electrical and Electronics Engineers fuzzy system transactions of expression, 2017. (Jing Gu,
Licheng Jiao,etc.Fuzzy Double C-Means Clustering Based on Sparse Self-
Representation.IEEE Trans Fuzzy Syst., 2017)) etc..
Above-mentioned innovatory algorithm all achieves certain improvement in terms of noise immunity and segmentation result, but there is also certain to ask
Topic.Above-mentioned algorithm introduces neighborhood space information by different modes, but has ignored geological information, while the neighbour utilized
Domain information is also only local neighborhood information.In order to more make full use of neighborhood information and geological information, symmetry letter is being considered
On the basis of breath, the present invention proposes a kind of method for extracting infrared target using Intuitionistic Fuzzy Clustering algorithm is improved.
【The content of the invention】
1st, purpose:Fuzzy clustering method has been widely applied in image segmentation field, but since the algorithm does not consider
The shortcomings that spatial information is to noise-sensitive, therefore tend not to obtain correctly cluster knot for noisy image fuzzy clustering algorithm
Fruit.And infrared image is there are contrast is low, the problems such as noise jamming, therefore fuzzy clustering is applied to image segmentation and can not often obtain
To correct segmentation result.For this problem, local region information and non local symmetric information are introduced into mould by the present invention
It pastes in C mean algorithms, so as to complete the segmentation and extraction to infrared target.
Segmentation for target in infrared image and extraction problem, the present invention, which proposes, a kind of to be gathered using improving intuitionistic fuzzy
The method that class algorithm extracts infrared target.On the one hand, this method using regional area degree of membership information and area classification information with
And symmetry information architecture weight coefficient, to the different weights of different classes of imparting of same pixel;On the other hand, in target letter
The regular distance item apart from regular terms and based on symmetric information based on area information is added in number, regular distance item embodies
Region average gray and the similitude of symmetrical pixels point and cluster centre.The innovatory algorithm of the present invention has taken into full account regional area
As a result information and non local symmetry information, the segmentation for target in infrared image achieve apparent improvement.
2nd, technical solution:In order to realize this purpose, technical scheme is as follows, true first with conspicuousness algorithm
Determine approximate region of the infrared target in infrared image.Improved Intuitionistic Fuzzy Clustering algorithm is recycled to cluster the region
Segmentation.Finally, nontarget area is rejected by simply post-processing.Taken into full account in innovatory algorithm local region information and
Non local symmetry information, therefore segmentation result has obtained apparent improvement.
The present invention is that a kind of utilization improves the method that Intuitionistic Fuzzy Clustering algorithm extracts infrared target, this method specific steps
It is as follows:
Step 1:Determine target approximate region.The notable figure of infrared image is obtained using conspicuousness algorithm, by notable
Figure obtains target approximate location into row threshold division.
Wherein, step 1 specifically comprises the following steps:
(1) is generated using Hou Xiaodi and Jonathan Halley that et al. the conspicuousness method based on image signatures proposed
The notable figure of infrared image is (referring to document:Hou Xiaodi, Jonathan Halley that, Christoffer kock image signatures:It is prominent
Go out sparse marking area American Institute of Electrical and Electronics Engineers pattern analysis and machine intelligence transactions .34 (1) 2012,194-
201.(X.D.Hou,J.Harel,C.Koch,Image signature:highlighting sparse salient
Regions, IEEE Trans.Pattern Anal.Mach.Intell.34 (1) (2012) 194-201.)), this method extraction
Gradient operator used is Sobel operators during notable figure.
(2) splits notable figure using Otsu threshold method, and segmentation result is bianry image.Non-zero region is recognized in segmentation result
It is set to target region.
Step 2:Using Ladakh Li Shinaa into up to et al. propose SLIC super-pixel segmentations method to target area
Domain carries out super-pixel segmentation, generates multiple zonules (referring to document:The .SLIC super-pixel .EPFL technology reports such as Ladakh Li Shina
Accuse .2010. (Achanta R, Shaji A, Smith K, et al.SLIC superpixels.Epfl, 2010.)).
Step 3:The mirror symmetry detection method based on registration proposed using Marcelo Xi Keen et al. completes target
Symmetrical shaft detection is (referring to document:Marcelo Xi Keen etc., the mirror symmetry based on registration detect, ICCV, 2017 (Marcelo
Cicconet,David G.C.Hildebrand,Hunter Elliott,Finding Mirror Symmetry via
Registration,ICCV,2017))。
Step 4:Cluster segmentation is carried out to infrared image using improved Intuitionistic Fuzzy Clustering algorithm.It is improved fuzzy poly-
Class algorithm object function is as follows:
Wherein, αjRepresent pixel xjWith its region RjAverage gray value similitude, xjWith region RjGray scale
It is worth closer, then αjIt is bigger, otherwise αjIt is smaller;βjRepresent pixel xjWith its symmetrical pixels point xjmGray value similitude, xj
With xjmGray value is closer, then illustrates that symmetry is better, then βjIt is bigger, otherwise, βjIt is smaller.αjAnd βjCalculation formula it is as follows:
Wherein, σ is the variance of image intensity value,For region RjThe average gray value of interior all pixels point,For xjNeighbour
The average gray value of domain pixel,For symmetrical pixels point xjmThe average gray value of all pixels point in neighborhood.
dIFS(xj,vi)、And dIFS(xjm,vi) it is Intuitionistic Fuzzy Distances, it is defined as follows:
dIFS(xj,vi)=(μ (xj)-μ(vi))2+(υ(xj)-υ(vi))2+(π(xj)-π(vi))2(4)
μ(xj), υ (xj), π (xj) it is respectively pixel x in Intuitionistic Fuzzy Clustering algorithmjDegree of membership, non-affiliated degree and still
Henan degree;Respectively region R in Intuitionistic Fuzzy Clustering algorithmjDegree of membership, non-affiliated degree and still
Henan degree;Symmetrical pixels point respectively in Intuitionistic Fuzzy Clustering algorithmDegree of membership, non-affiliated
Degree and hesitation degree.Its calculation formula difference is as follows:
π(xj)=1- μ (xj)-υ(xj)(9)
μ(vi), υ (vi), π (vi) it is respectively cluster centre viDegree of membership, non-affiliated degree and hesitation degree.
Each symbol definition is as follows in object function and above-mentioned calculation formula:N is pixel sum;J sits for pixel position
Mark;C is classification number;I is classification ordinal number;uijIt is j-th of pixel compared with the degree of membership of the i-th class;viIn cluster for the i-th class
The heart;μ(vi) it is cluster centre viDegree of membership;υ(vi) it is cluster centre viNon-affiliated degree;π(vi) it is cluster centre viStill
Henan degree;M is fuzzy factor;WijFor weight coefficient;xjFor the pixel value of j-th of pixel;μ(xj) it is xjDegree of membership;υ(xj)
For xjNon-affiliated degree;π(xj) it is xjHesitation degree;RjFor xjPlace zonule;For xjThe average gray of place zonule
Value;ForDegree of membership;ForNon-affiliated degree;ForHesitation degree;xjmFor xjSymmetric points;ForDegree of membership;ForNon-affiliated degree;ForHesitation degree;xminIt is minimum for gradation of image
Value;xmaxFor gradation of image maximum;For minimum gray value in all areas;For gray scale in all areas most
Big value;For the minimum gray value of all symmetrical pixels points;For the gray scale maximum of all symmetrical pixels points;αj
For an adaptive constant, pixel x is representedjWith its region RjAverage gray value similitude;βjIt is adaptive normal for one
Number represents pixel xjWith its symmetrical pixels point xjmGray value similitude;For xjNeighborhood territory pixel point average gray
Value;λ is constant.
Weight coefficient WijCalculating process it is as follows:
RijIt is constrained for area information, SijFor symmetric constraints, calculating process is as follows:
Wherein,Represent pixel xjThe region R at placejAverage membership, LiIt represents in region RjIn belong to the i-th class
Pixel number, L represents region RjIn total pixel number.For xjNeighborhood territory pixel point average gray value.
γjRepresentation space distance restraint, pixel distance objective center is more remote, then γjIt is smaller, in pixel distance objective
The heart is nearer, then γjIt is bigger.Its calculating process is as follows:
Wherein (pj,qj) represent xjSpace coordinates, (p0,q0) space coordinates of target's center is represented, a represents elliptical length
Axis, b represent elliptical short axle.ξ is a small constant, and 0.2 is set in the present invention.
Tertiary target is included in experimental data used in the present invention:Infrared pedestrian, infrared ship, infrared aircraft.According to three
The shape of kind target, the present invention are fitted the shape of target using elliptical form so that the pixel within ellipse
Point has larger γj, and the pixel outside oval has smaller γj, and distance objective center is nearer, γjMore
Greatly.Symmetry axis position is determined as to the row coordinate q of the ordinate of target's center, i.e. elliptical center0.In infrared image,
Target should be highlight regions, therefore the grey scale change trend of symmetry axis column is usually " trough-wave crest-trough ", takes wave crest
Length is as elliptical long axis a.Similarly, for every a line of target area, grey scale change trend is also usually " trough-ripple
Peak-trough " takes in all rows maximum wave crest length as elliptical short axle b, row coordinate of the row coordinate as elliptical center
p0。
Using lagrange's method of multipliers, according to object function derivation, degree of membership u can be derivedijWith cluster centre viPerson in servitude
Category degree μ (vi), non-affiliated degree υ (vi), hesitation degree π (vi) iterative formula:
Wherein vkRepresent the cluster centre of kth class.
The step of improvement Intuitionistic Fuzzy Clustering algorithm proposed by the present invention is split for image be:
(1) sets classification number c, maximum iteration T, iteration stopping threshold epsilon.
(2) each element u in random initializtions subordinated-degree matrix UijWith each cluster centre viDegree of membership μ (vi)、
Non-affiliated degree υ (vi), hesitation degree π (vi)。
(3) calculates each element W in weight coefficient matrix W according to formula (16), (17), (18), (19)ij。
(4) updates each element u in subordinated-degree matrix U according to formula (20)ij。
(5) calculates cluster centre v respectively according to formula (21), (22), (23)iDegree of membership μ (vi), non-affiliated degree υ
(vi), hesitation degree π (vi)。
(6) if | U (t+1)-U (t) | < ε (t is iterations) or iterations stop more than maximum iteration T
Iteration carries out step (6);Otherwise, step (2) is carried out.
(7) defuzzifications complete image segmentation.
Wherein, c is the classification number in fuzzy clustering, need to manually be set;Maximum iteration T and iteration stopping threshold epsilon are used
In control iterative process, then stop less than threshold epsilon more than the T or front and rear degree of membership differences that iteration obtains twice when iterations
Iterative process.
Subordinated-degree matrix U=(uij)m×n, it is one and an equal amount of matrix of image, m, n are the length and width of image, in U
Element be degree of membership uij;Random initializtion subordinated-degree matrix refers to each element u in subordinated-degree matrix UijIt is random to assign
Value, while make u after assignmentijMeet following condition:
Similarly, weight coefficient matrix W=(Wij)m×n, it is one and an equal amount of matrix of image, the element in W is power
Weight coefficient Wij;Update weight coefficient matrix and subordinated-degree matrix refer to according to WijAnd uijCalculation formula calculate new WijWith
uij。
Defuzzification refers to determine each pixel x according to the subordinated-degree matrix U that above-mentioned (1)-(6) step obtainsj's
The process of generic, pixel xjClassification determine according to the following formula:
Wherein i is classification sequence number, for integer, i=1,2 ... c.Then pixel xjClassification be uijIt is right to obtain maximum when institute
The classification sequence number answered.Subordinated-degree matrix according to obtaining determines that the process of each pixel classification is defuzzification process.
Step 5:The image segmentation result that step 4 obtains be bianry image, prospect 1, background 0.But before obtaining
Nontarget area may be included in scape.In order to extract infrared target, it is necessary to the image segmentation result that step 4 obtains into rear
Continuous processing, rejects nontarget area.Subsequent processing is included according to herein below:
(1) region areas are less than some threshold value, then reject the region.
(2) rejects nontarget area according to the length-width ratio in region.
(3) region that rejectings are connected with image boundary
In step 5, the aspect ratio range of infrared pedestrian is set to 1~4 by the present invention, and infrared ship length-width ratio is set to 1~
10, infrared aircraft length-width ratio is set to 1~3.
3rd, advantage and effect:Traditional FCM Algorithms do not consider spatial information and geological information, it is impossible to well
Noisy image is handled, correct segmentation result can not be obtained for noisy image.And infrared image is low with contrast, module of boundary
Paste, the problems such as containing noise, therefore traditional fuzzy C mean algorithms are used to often can not preferably be divided during infrared Image Segmentation
Cut result.The local region information of infrared image and non local symmetrical letter have been taken into full account in innovatory algorithm proposed by the present invention
Breath, the geometrically symmetric information of area grayscale information and target is attached in intuition FCM algorithms, improves intuition FCM algorithms
Segmentation result.Meanwhile it is highlight regions that infrared target is considered in innovatory algorithm in infrared image, symmetry axis should be located at mesh
This two prioris on mark.Innovatory algorithm achieves good segmentation result in experimental data.In innovatory algorithm fully
Consider area grayscale information and degree of membership information and non local symmetric information, therefore being capable of preferable Ground Split infrared target.
Have a vast market prospect and application value.
【Description of the drawings】
Fig. 1 is the Method And Principle block diagram that the present invention extracts infrared target using improved Intuitionistic Fuzzy Clustering algorithm.
Fig. 2 a are present invention determine that the infrared pedestrian of pedestrian area schemes in infrared image.
Fig. 2 b be present invention determine that in infrared image naval vessel region infrared ship figure.
Fig. 2 c are present invention determine that the infrared aircraft figure in aircraft region in infrared image.
Fig. 2 d are the corresponding notable figure of the infrared pedestrian image.
Fig. 2 e are the corresponding notable figure of the infrared ship image.
Fig. 2 f are the corresponding notable figure of the infrared aircraft brake disc.
It is final infrared pedestrian target region that the present invention is determined using conspicuousness that Fig. 2 g, which are, i.e., region to be split.
The final infrared ship target region that Fig. 2 h are determined for the present invention using conspicuousness, i.e., region to be split.
Fig. 2 i determine final infrared Aircraft Targets region, i.e., region to be split for the present invention using conspicuousness.
Fig. 3 a are the result of infrared pedestrian's super-pixel segmentation.
Fig. 3 b are the result of infrared ship super-pixel segmentation.
Fig. 3 c are the result of infrared aircraft super-pixel segmentation.
Fig. 4 a are the result of the symmetrical shaft detection of infrared pedestrian.
Fig. 4 b are the result of the symmetrical shaft detection of infrared ship.
Fig. 4 c are the result of the symmetrical shaft detection of infrared aircraft.
Fig. 5 a are the segmentation result that improvement Intuitionistic Fuzzy Clustering algorithm proposed by the present invention is used for infrared pedestrian image.
Fig. 5 b are the segmentation result that improvement Intuitionistic Fuzzy Clustering algorithm proposed by the present invention is used for infrared ship image.
Fig. 5 c are the segmentation result that improvement Intuitionistic Fuzzy Clustering algorithm proposed by the present invention is used for infrared aircraft brake disc.
Fig. 6 a are final result of the present invention for the extraction of infrared pedestrian target.
Fig. 6 b are final result of the present invention for infrared ship target extraction.
Fig. 6 c are final result of the present invention for the extraction of infrared Aircraft Targets.
Fig. 7 a are the artworks that the present invention is used to extract infrared pedestrian.
Fig. 7 b are the artworks that the present invention is used to extract infrared ship.
Fig. 7 c are the artworks that the present invention is used to extract infrared aircraft.
Fig. 7 d present invention is used to extract the final result of infrared pedestrian.
Fig. 7 e present invention is used to extract the final result of infrared ship.
Fig. 7 f present invention is used to extract the final result of infrared aircraft.
【Specific embodiment】
Technical solution for a better understanding of the present invention is made embodiments of the present invention below in conjunction with attached drawing further
Description.The experimental data that the present invention uses includes 265 infrared images, is divided into infrared pedestrian, infrared ship, three kinds of infrared aircraft
Type.Embodiment of the present invention is described in detail using infrared pedestrian as example below.
The principle of the present invention block diagram is as shown in Figure 1, the specific implementation step of the present invention is as follows:
Step 1:Determine infrared pedestrian's approximate region.The notable figure of infrared image is obtained using conspicuousness algorithm, by right
Notable figure obtains infrared pedestrian's approximate location into row threshold division.
Wherein, step 1 specifically comprises the following steps:
(1) is generated using Hou Xiaodi and Jonathan Halley that et al. the conspicuousness method based on image signatures proposed
The notable figure of infrared image is (referring to document:Hou Xiaodi, Jonathan Halley that, Christoffer kock image signatures:It is prominent
Go out sparse marking area American Institute of Electrical and Electronics Engineers pattern analysis and machine intelligence transactions .34 (1) 2012,194-
201.(X.D.Hou,J.Harel,C.Koch,Image signature:highlighting sparse salient
Regions, IEEE Trans.Pattern Anal.Mach.Intell.34 (1) (2012) 194-201.)), this method extraction
Gradient operator used is Sobel operators during notable figure.
(2) splits notable figure using Otsu threshold method, and segmentation result is bianry image.Non-zero region is recognized in segmentation result
It is set to pedestrian region.
Fig. 2 a are present invention determine that the infrared pedestrian of pedestrian area schemes in infrared image;Fig. 2 d are the infrared pedestrian image pair
The notable figure answered;It is final infrared pedestrian target region that the present invention is determined using conspicuousness that Fig. 2 g, which are, i.e., region to be split.
Fig. 2 b be present invention determine that in infrared image naval vessel region infrared ship figure;Fig. 2 e are the infrared ship image pair
The notable figure answered;The final infrared ship target region that Fig. 2 h are determined for the present invention using conspicuousness, i.e., region to be split.
Fig. 2 c are present invention determine that the infrared aircraft figure in aircraft region in infrared image.Fig. 2 f correspond to for the infrared aircraft brake disc
Notable figure.Fig. 2 i determine final infrared Aircraft Targets region, i.e., region to be split for the present invention using conspicuousness.
Step 2:Using Ladakh Li Shinaa into up to et al. propose SLIC super-pixel segmentations method to target area
Domain carries out super-pixel segmentation, generates multiple zonules (referring to document:The .SLIC super-pixel .EPFL technology reports such as Ladakh Li Shina
Accuse .2010. (Achanta R, Shaji A, Smith K, et al.SLIC superpixels.Epfl, 2010.)).
Fig. 3 a are the result of infrared pedestrian's super-pixel segmentation.Fig. 3 b are the result of infrared ship super-pixel segmentation.Fig. 3 c are
The result of infrared aircraft super-pixel segmentation.
Step 3:The mirror symmetry detection method based on registration proposed using Marcelo Xi Keen et al. completes target
Symmetrical shaft detection is (referring to document:Marcelo Xi Keen etc., the mirror symmetry based on registration detect, ICCV, 2017 (Marcelo
Cicconet,David G.C.Hildebrand,Hunter Elliott,Finding Mirror Symmetry via
Registration,ICCV,2017))。
Fig. 4 a are the result of the symmetrical shaft detection of infrared pedestrian.Fig. 4 b are the result of the symmetrical shaft detection of infrared ship.Fig. 4 c are
The result of the infrared symmetrical shaft detection of aircraft.
Step 4:Cluster segmentation is carried out to infrared image using improved Intuitionistic Fuzzy Clustering algorithm.It is improved fuzzy poly-
Class algorithm object function is as follows:
Wherein, αjRepresent pixel xjWith its region RjAverage gray value similitude, xjWith region RjGray scale
It is worth closer, then αjIt is bigger, otherwise αjIt is smaller;βjRepresent pixel xjWith its symmetrical pixels point xjmGray value similitude, xj
With xjmGray value is closer, then illustrates that symmetry is better, then βjIt is bigger, otherwise, βjIt is smaller.αjAnd βjCalculation formula it is as follows:
Wherein, σ is the variance of image intensity value,For region RjThe average gray value of interior all pixels point,For xj's
The average gray value of neighborhood territory pixel point,For symmetrical pixels point xjmThe average gray value of all pixels point in neighborhood.
dIFS(xj,vi)、And dIFS(xjm,vi) it is Intuitionistic Fuzzy Distances, it is defined as follows:
μ(xj), υ (xj), π (xj) it is respectively pixel x in Intuitionistic Fuzzy Clustering algorithmjDegree of membership, non-affiliated degree and still
Henan degree;Respectively region R in Intuitionistic Fuzzy Clustering algorithmjDegree of membership, non-affiliated degree and still
Henan degree;Symmetrical pixels point respectively in Intuitionistic Fuzzy Clustering algorithmDegree of membership, non-person in servitude
Category degree and hesitation degree.Its calculation formula difference is as follows:
π(xj)=1- μ (xj)-υ(xj) (33)
μ(vi), υ (vi), π (vi) it is respectively cluster centre viDegree of membership, non-affiliated degree and hesitation degree.
Each symbol definition is as follows in object function and above-mentioned calculation formula:N is pixel sum;J sits for pixel position
Mark;C is classification number;I is classification ordinal number;uijIt is j-th of pixel compared with the degree of membership of the i-th class;viIn cluster for the i-th class
The heart;μ(vi) it is cluster centre viDegree of membership;υ(vi) it is cluster centre viNon-affiliated degree;π(vi) it is cluster centre viStill
Henan degree;M is fuzzy factor;WijFor weight coefficient;xjFor the pixel value of j-th of pixel;μ(xj) it is xjDegree of membership;υ(xj)
For xjNon-affiliated degree;π(xj) it is xjHesitation degree;RjFor xjPlace zonule;For xjThe average gray value of place zonule;ForDegree of membership;ForNon-affiliated degree;ForHesitation degree;xjmFor xjSymmetric points;ForDegree of membership;ForNon-affiliated degree;ForHesitation degree;xminIt is minimum for gradation of image
Value;xmaxFor gradation of image maximum;For minimum gray value in all areas;For gray scale in all areas most
Big value;For the minimum gray value of all symmetrical pixels points;For the gray scale maximum of all symmetrical pixels points;αj
For an adaptive constant, pixel x is representedjWith its region RjAverage gray value similitude;βjIt is adaptive normal for one
Number represents pixel xjWith its symmetrical pixels point xjmGray value similitude;For xjNeighborhood territory pixel point average gray
Value;λ is constant.
Weight coefficient WijCalculating process it is as follows:
RijIt is constrained for area information, SijFor symmetric constraints, calculating process is as follows:
Wherein,Represent pixel xjThe region R at placejAverage membership, LiIt represents in region RjIn belong to the i-th class
Pixel number, L represents region RjIn total pixel number.For xjNeighborhood territory pixel point average gray value.
γjRepresentation space distance restraint, pixel distance objective center is more remote, then γjIt is smaller, in pixel distance objective
The heart is nearer, then γjIt is bigger.Its calculating process is as follows:
Wherein (pj,qj) represent xjSpace coordinates, (p0,q0) space coordinates of target's center is represented, a represents elliptical length
Axis, b represent elliptical short axle.ξ is a small constant, and 0.2 is set in the present invention.
According to the shape of infrared pedestrian, the present invention is fitted the shape of infrared pedestrian using elliptical form so that
Pixel within ellipse has larger γj, and the pixel outside oval has smaller γj, and apart from mesh
Mark center is nearer, γjIt is bigger.Symmetry axis position is determined as to the row coordinate of the ordinate at pedestrian center, i.e. elliptical center
q0.In infrared image, pedestrian should be highlight regions, therefore the grey scale change trend of symmetry axis column is usually " trough-ripple
Peak-trough " takes wave crest length as elliptical long axis a.Similarly, for every a line of pedestrian area, grey scale change trend
Usually " trough-wave crest-trough " takes maximum wave crest length in all rows to be used as oval as elliptical short axle b, the row coordinate
The row coordinate p at center0。
Using lagrange's method of multipliers, according to object function derivation, degree of membership u can be derivedijWith cluster centre viPerson in servitude
Category degree μ (vi), non-affiliated degree υ (vi), hesitation degree π (vi) iterative formula:
Wherein vkRepresent the cluster centre of kth class.
The step of improvement Intuitionistic Fuzzy Clustering algorithm proposed by the present invention is split for image be:
(1) sets classification number c, maximum iteration T, iteration stopping threshold epsilon.
(2) each element u in random initializtions subordinated-degree matrix UijWith each cluster centre viDegree of membership μ (vi)、
Non-affiliated degree υ (vi), hesitation degree π (vi)。
(3) calculates each element W in weight coefficient matrix W according to formula (40), (41), (42), (43)ij。
(4) updates each element u in subordinated-degree matrix U according to formula (44)ij。
(5) calculates cluster centre v respectively according to formula (45), (46), (47)iDegree of membership μ (vi), non-affiliated degree υ
(vi), hesitation degree π (vi)。
(6) if | U (t+1)-U (t) | < ε (t is iterations) or iterations stop more than maximum iteration T
Iteration carries out step (6);Otherwise, step (2) is carried out.
(7) defuzzifications complete image segmentation.
Wherein, c is the classification number in fuzzy clustering, need to manually be set;In the present invention, infrared pedestrian image is divided into 4 classes,
C=4 in i.e. infrared pedestrian image;Infrared ship image is divided into 3 classes, i.e. c=3 in infrared ship image;By infrared aircraft figure
As being divided into 2 classes, i.e., c=2 in infrared aircraft brake disc.
Maximum iteration T and iteration stopping threshold epsilon are for controlling iterative process, when iterations is more than T or front and rear two
The degree of membership difference that secondary iteration obtains is less than threshold epsilon, then stops iterative process.T=100 in the present invention, ε=10-5, that is, work as iteration
When number is more than 100 times, iterative process stops;Or when the difference of front and rear iteration degree of membership twice is less than 10-5When, iteration is similary
Stop.
Subordinated-degree matrix U=(uij)m×n, it is one and an equal amount of matrix of image, m, n are the length and width of image, in U
Element be degree of membership uij;Random initializtion subordinated-degree matrix refers to each element u in subordinated-degree matrix UijIt is random to assign
Value, while make u after assignmentijMeet following condition:
Similarly, weight coefficient matrix W=(Wij)m×n, it is one and an equal amount of matrix of image, the element in W is power
Weight coefficient Wij;Update weight coefficient matrix and subordinated-degree matrix refer to according to WijAnd uijCalculation formula calculate new WijWith
uij。
Defuzzification refers to determine each pixel x according to the subordinated-degree matrix U that above-mentioned (1)-(6) step obtainsj's
The process of generic, pixel xjClassification determine according to the following formula:
Wherein i is classification sequence number, for integer, i=1,2 ... c.Then pixel xjClassification be uijIt is right to obtain maximum when institute
The classification sequence number answered.Subordinated-degree matrix according to obtaining determines that the process of each pixel classification is defuzzification process.
Fig. 5 a are the segmentation result that improvement Intuitionistic Fuzzy Clustering algorithm proposed by the present invention is used for infrared pedestrian image;Fig. 5 b
The segmentation result of infrared ship image is used for for improvement Intuitionistic Fuzzy Clustering algorithm proposed by the present invention;Fig. 5 c propose for the present invention
The segmentation result for improving Intuitionistic Fuzzy Clustering algorithm and being used for infrared aircraft brake disc.
Step 5:The image segmentation result that step 4 obtains be bianry image, prospect 1, background 0.But before obtaining
Nontarget area may be included in scape.In order to extract infrared target, it is necessary to the image segmentation result that step 4 obtains into rear
Continuous processing, rejects nontarget area.Subsequent processing is included according to herein below:
(1) region areas are less than some threshold value, then reject the region.
(2) rejects nontarget area according to the length-width ratio in region.
(3) region that rejectings are connected with image boundary
In step 5, region area threshold value is set to 100 by the present invention, and the aspect ratio range of infrared pedestrian is set to 1~4,
Infrared ship length-width ratio is set to 1~10, and infrared aircraft length-width ratio is set to 1~3.Therefore, area of the area less than 100 in step 5
Domain will be removed, and the region being connected with image boundary will be removed, and not met the region of target length-width ratio and will be removed.
Fig. 6 a are final result of the present invention for the extraction of infrared pedestrian target;Fig. 6 b are used for infrared ship mesh for the present invention
Mark the final result of extraction;Fig. 6 c are final result of the present invention for the extraction of infrared Aircraft Targets.
In order to show the effect of the present invention, Fig. 7 a are the artworks that the present invention is used to extract infrared pedestrian.Fig. 7 b are the present invention
For extracting the artwork of infrared ship.Fig. 7 c are the artworks that the present invention is used to extract infrared aircraft.Fig. 7 d present invention is used to extract
The final result of infrared pedestrian.Fig. 7 e present invention is used to extract the final result of infrared ship.Fig. 7 f present invention is red for extracting
The final result of outer aircraft.
Experimental data of the present invention includes three kinds of targets:Infrared pedestrian, infrared ship and infrared aircraft.The present invention
The local region information of infrared image and non local symmetric information are attached in Intuitionistic Fuzzy Clustering by the innovatory algorithm of proposition,
Two prioris on target, therefore energy should be located at for highlight regions, symmetry axis by also using in infrared image target simultaneously
Enough effectively improve the segmentation result of fuzzy clustering.As can be seen from Figure 7, it is equal for three kinds of target innovatory algorithms proposed by the present invention
Good segmentation result can be obtained.
Claims (7)
1. a kind of utilize the method for improving Intuitionistic Fuzzy Clustering algorithm extraction infrared target, it is characterised in that:This method specifically walks
It is rapid as follows:
Step 1:Determine target approximate region;The notable figure of infrared image is obtained using conspicuousness algorithm, by notable figure into
Row threshold division obtains target approximate location;
Step 2:Using Ladakh Li Shinaa into up to et al. propose SLIC super-pixel segmentations method to target area into
Row super-pixel segmentation generates multiple zonules;
Step 3:It is symmetrical that the mirror symmetry detection method based on registration proposed using Marcelo Xi Keen et al. completes target
Shaft detection;
Step 4:Cluster segmentation is carried out to infrared image using improved Intuitionistic Fuzzy Clustering algorithm;Improved fuzzy clustering is calculated
Method object function is as follows:
<mrow>
<mi>J</mi>
<mo>=</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>c</mi>
</munderover>
<msub>
<mi>W</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<msubsup>
<mi>u</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
<mi>m</mi>
</msubsup>
<mrow>
<mo>(</mo>
<msup>
<mi>d</mi>
<mrow>
<mi>I</mi>
<mi>F</mi>
<mi>S</mi>
</mrow>
</msup>
<mo>(</mo>
<mrow>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>,</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
</mrow>
<mo>)</mo>
<mo>+</mo>
<msub>
<mi>&alpha;</mi>
<mi>j</mi>
</msub>
<msup>
<mi>d</mi>
<mrow>
<mi>I</mi>
<mi>F</mi>
<mi>S</mi>
</mrow>
</msup>
<mo>(</mo>
<mrow>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>,</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
</mrow>
<mo>)</mo>
<mo>+</mo>
<msub>
<mi>&beta;</mi>
<mi>j</mi>
</msub>
<msup>
<mi>d</mi>
<mrow>
<mi>I</mi>
<mi>F</mi>
<mi>S</mi>
</mrow>
</msup>
<mo>(</mo>
<mrow>
<msubsup>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
<mo>-</mo>
</msubsup>
<mo>,</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein, αjRepresent pixel xjWith its region RjAverage gray value similitude, xjWith region RjGray value get over
It approaches, then αjIt is bigger, otherwise αjIt is smaller;βjRepresent pixel xjWith its symmetrical pixels point xjmGray value similitude, xjWith xjm
Gray value is closer, then illustrates that symmetry is better, then βjIt is bigger, otherwise, βjIt is smaller;αjAnd βjCalculation formula it is as follows:
<mrow>
<msub>
<mi>&alpha;</mi>
<mi>j</mi>
</msub>
<mo>=</mo>
<mi>exp</mi>
<mrow>
<mo>(</mo>
<mo>-</mo>
<mfrac>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mrow>
<mn>2</mn>
<msup>
<mi>&sigma;</mi>
<mn>2</mn>
</msup>
</mrow>
</mfrac>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>&beta;</mi>
<mi>j</mi>
</msub>
<mo>=</mo>
<mi>exp</mi>
<mrow>
<mo>(</mo>
<mo>-</mo>
<mfrac>
<msup>
<mrow>
<mo>(</mo>
<mover>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>-</mo>
</mover>
<mo>-</mo>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>-</mo>
</mover>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mrow>
<mn>2</mn>
<msup>
<mi>&sigma;</mi>
<mn>2</mn>
</msup>
</mrow>
</mfrac>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>3</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein, σ is the variance of image intensity value,For region RjThe average gray value of interior all pixels point,For xjNeighborhood picture
The average gray value of vegetarian refreshments,For symmetrical pixels point xjmThe average gray value of all pixels point in neighborhood;
dIFS(xj,vi)、And dIFS(xjm,vi) it is Intuitionistic Fuzzy Distances, it is defined as follows:
dIFS(xj,vi)=(μ (xj)-μ(vi))2+(υ(xj)-υ(vi))2+(π(xj)-π(vi))2 (4)
<mrow>
<msup>
<mi>d</mi>
<mrow>
<mi>I</mi>
<mi>F</mi>
<mi>S</mi>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>,</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msup>
<mrow>
<mo>(</mo>
<mi>&mu;</mi>
<mo>(</mo>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>)</mo>
<mo>-</mo>
<mi>&mu;</mi>
<mo>(</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<mi>&upsi;</mi>
<mo>(</mo>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>)</mo>
<mo>-</mo>
<mi>&upsi;</mi>
<mo>(</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<mi>&pi;</mi>
<mo>(</mo>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>)</mo>
<mo>-</mo>
<mi>&pi;</mi>
<mo>(</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>5</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msup>
<mi>d</mi>
<mrow>
<mi>I</mi>
<mi>F</mi>
<mi>S</mi>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>-</mo>
</mover>
<mo>,</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msup>
<mrow>
<mo>(</mo>
<mi>&mu;</mi>
<mo>(</mo>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>-</mo>
</mover>
<mo>)</mo>
<mo>-</mo>
<mi>&mu;</mi>
<mo>(</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<mi>&upsi;</mi>
<mo>(</mo>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>-</mo>
</mover>
<mo>)</mo>
<mo>-</mo>
<mi>&upsi;</mi>
<mo>(</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<mi>&pi;</mi>
<mo>(</mo>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>-</mo>
</mover>
<mo>)</mo>
<mo>-</mo>
<mi>&pi;</mi>
<mo>(</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>6</mn>
<mo>)</mo>
</mrow>
</mrow>
μ(xj), υ (xj), π (xj) it is respectively pixel x in Intuitionistic Fuzzy Clustering algorithmjDegree of membership, non-affiliated degree and hesitation degree;Respectively region R in Intuitionistic Fuzzy Clustering algorithmjDegree of membership, non-affiliated degree and hesitation degree;Symmetrical pixels point respectively in Intuitionistic Fuzzy Clustering algorithmDegree of membership, non-affiliated degree and
Hesitation degree;Its calculation formula difference is as follows:
<mrow>
<mi>&mu;</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mi>min</mi>
</msub>
</mrow>
<mrow>
<msub>
<mi>x</mi>
<mrow>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mi>min</mi>
</msub>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>7</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>&upsi;</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<mn>1</mn>
<mo>-</mo>
<mi>&mu;</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mn>1</mn>
<mo>+</mo>
<mi>&lambda;</mi>
<mi>&mu;</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>8</mn>
<mo>)</mo>
</mrow>
</mrow>
π(xj)=1- μ (xj)-υ(xj) (9)
<mrow>
<mi>&mu;</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>-</mo>
<msub>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>)</mo>
</mrow>
<mi>min</mi>
</msub>
</mrow>
<mrow>
<msub>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>)</mo>
</mrow>
<mi>max</mi>
</msub>
<mo>-</mo>
<msub>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>)</mo>
</mrow>
<mi>min</mi>
</msub>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>10</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>&upsi;</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<mn>1</mn>
<mo>-</mo>
<mi>&mu;</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mn>1</mn>
<mo>+</mo>
<mi>&lambda;</mi>
<mi>&mu;</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>11</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>&pi;</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mn>1</mn>
<mo>-</mo>
<mi>&mu;</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mi>&upsi;</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>12</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>&mu;</mi>
<mrow>
<mo>(</mo>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>-</mo>
</mover>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>-</mo>
<msub>
<mrow>
<mo>(</mo>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>-</mo>
</mover>
<mo>)</mo>
</mrow>
<mi>min</mi>
</msub>
</mrow>
<mrow>
<msub>
<mrow>
<mo>(</mo>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>-</mo>
</mover>
<mo>)</mo>
</mrow>
<mi>max</mi>
</msub>
<mo>-</mo>
<msub>
<mrow>
<mo>(</mo>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>-</mo>
</mover>
<mo>)</mo>
</mrow>
<mi>min</mi>
</msub>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>13</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>&upsi;</mi>
<mrow>
<mo>(</mo>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>-</mo>
</mover>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<mn>1</mn>
<mo>-</mo>
<mi>&mu;</mi>
<mrow>
<mo>(</mo>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>-</mo>
</mover>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mn>1</mn>
<mo>+</mo>
<mi>&lambda;</mi>
<mi>&mu;</mi>
<mrow>
<mo>(</mo>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>-</mo>
</mover>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>14</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>&pi;</mi>
<mrow>
<mo>(</mo>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>-</mo>
</mover>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mn>1</mn>
<mo>-</mo>
<mi>&mu;</mi>
<mrow>
<mo>(</mo>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>-</mo>
</mover>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mi>&upsi;</mi>
<mrow>
<mo>(</mo>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>-</mo>
</mover>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>15</mn>
<mo>)</mo>
</mrow>
</mrow>
μ(vi), υ (vi), π (vi) it is respectively cluster centre viDegree of membership, non-affiliated degree and hesitation degree;
Each symbol definition is as follows in object function and above-mentioned calculation formula:N is pixel sum;J is pixel position coordinates;c
For classification number;I is classification ordinal number;uijIt is j-th of pixel compared with the degree of membership of the i-th class;viFor the cluster centre of the i-th class;μ
(vi) it is cluster centre viDegree of membership;υ(vi) it is cluster centre viNon-affiliated degree;π(vi) it is cluster centre viHesitation degree;
M is fuzzy factor;WijFor weight coefficient;xjFor the pixel value of j-th of pixel;μ(xj) it is xjDegree of membership;υ(xj) it is xj's
Non-affiliated degree;π(xj) it is xjHesitation degree;RjFor xjPlace zonule;For xjThe average gray value of place zonule;
ForDegree of membership;ForNon-affiliated degree;ForHesitation degree;xjmFor xjSymmetric points;ForDegree of membership;ForNon-affiliated degree;ForHesitation degree;xminFor gradation of image minimum value;xmaxFor
Gradation of image maximum;For minimum gray value in all areas;For gray scale maximum in all areas;For the minimum gray value of all symmetrical pixels points;For the gray scale maximum of all symmetrical pixels points;αjFor one
Adaptive constant represents pixel xjWith its region RjAverage gray value similitude;βjFor an adaptive constant, table
Show pixel xjWith its symmetrical pixels point xjmGray value similitude;For xjNeighborhood territory pixel point average gray value;λ is
Constant;
Weight coefficient WijCalculating process it is as follows:
<mrow>
<msub>
<mi>W</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>R</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>+</mo>
<msub>
<mi>S</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
</mrow>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>c</mi>
</munderover>
<msub>
<mi>R</mi>
<mrow>
<mi>k</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>+</mo>
<msub>
<mi>S</mi>
<mrow>
<mi>k</mi>
<mi>j</mi>
</mrow>
</msub>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>16</mn>
<mo>)</mo>
</mrow>
</mrow>
RijIt is constrained for area information, SijFor symmetric constraints, calculating process is as follows:
<mrow>
<msub>
<mi>R</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>=</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>-</mo>
<mi>exp</mi>
<mo>(</mo>
<mrow>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mrow>
<mn>1</mn>
<mo>-</mo>
<msub>
<mi>u</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>&times;</mo>
<mrow>
<mo>(</mo>
<mi>exp</mi>
<mo>(</mo>
<mrow>
<mo>-</mo>
<mfrac>
<msub>
<mi>L</mi>
<mi>i</mi>
</msub>
<mi>L</mi>
</mfrac>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>17</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein,Represent pixel xjThe region R at placejAverage membership, LiIt represents in region RjIn belong to the picture of the i-th class
The number of vegetarian refreshments, L represent region RjIn total pixel number;For xjNeighborhood territory pixel point average gray value;
γjRepresentation space distance restraint, pixel distance objective center is more remote, then γjSmaller, pixel distance objective center is got over
Closely, then γjIt is bigger;Its calculating process is as follows:
Wherein (pj,qj) represent xjSpace coordinates, (p0,q0) space coordinates of target's center is represented, a represents elliptical long axis, b
Represent elliptical short axle;ξ is a small constant, is set to 0.2;
Using lagrange's method of multipliers, according to object function derivation, degree of membership u can be derivedijWith cluster centre viDegree of membership μ
(vi), non-affiliated degree υ (vi), hesitation degree π (vi) iterative formula:
<mrow>
<msub>
<mi>u</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mfrac>
<mn>1</mn>
<mrow>
<msub>
<mi>W</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<msup>
<mi>d</mi>
<mrow>
<mi>I</mi>
<mi>F</mi>
<mi>S</mi>
</mrow>
</msup>
<mo>(</mo>
<mrow>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>,</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
</mrow>
<mo>)</mo>
<mo>+</mo>
<msup>
<mi>d</mi>
<mrow>
<mi>I</mi>
<mi>F</mi>
<mi>S</mi>
</mrow>
</msup>
<mo>(</mo>
<mrow>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>,</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
</mrow>
<mo>)</mo>
<mo>+</mo>
<msub>
<mi>&beta;</mi>
<mi>j</mi>
</msub>
<msup>
<mi>d</mi>
<mrow>
<mi>I</mi>
<mi>F</mi>
<mi>S</mi>
</mrow>
</msup>
<mo>(</mo>
<mrow>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>-</mo>
</mover>
<mo>,</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>c</mi>
</munderover>
<mfrac>
<mn>1</mn>
<mrow>
<msub>
<mi>W</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<msup>
<mi>d</mi>
<mrow>
<mi>I</mi>
<mi>F</mi>
<mi>S</mi>
</mrow>
</msup>
<mo>(</mo>
<mrow>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>,</mo>
<msub>
<mi>v</mi>
<mi>k</mi>
</msub>
</mrow>
<mo>)</mo>
<mo>+</mo>
<msup>
<mi>d</mi>
<mrow>
<mi>I</mi>
<mi>F</mi>
<mi>S</mi>
</mrow>
</msup>
<mo>(</mo>
<mrow>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>,</mo>
<msub>
<mi>v</mi>
<mi>k</mi>
</msub>
</mrow>
<mo>)</mo>
<mo>+</mo>
<msub>
<mi>&beta;</mi>
<mi>j</mi>
</msub>
<msup>
<mi>d</mi>
<mrow>
<mi>I</mi>
<mi>F</mi>
<mi>S</mi>
</mrow>
</msup>
<mo>(</mo>
<mrow>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>-</mo>
</mover>
<mo>,</mo>
<msub>
<mi>v</mi>
<mi>k</mi>
</msub>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>20</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>&mu;</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<msub>
<mi>W</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<msubsup>
<mi>u</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
<mi>m</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>&mu;</mi>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
<mo>+</mo>
<msub>
<mi>&alpha;</mi>
<mi>j</mi>
</msub>
<mi>&mu;</mi>
<mo>(</mo>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>)</mo>
<mo>+</mo>
<msub>
<mi>&beta;</mi>
<mi>j</mi>
</msub>
<mi>&mu;</mi>
<mo>(</mo>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>-</mo>
</mover>
<mo>)</mo>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<msub>
<mi>W</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<msubsup>
<mi>u</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
<mi>m</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>+</mo>
<msub>
<mi>&alpha;</mi>
<mi>j</mi>
</msub>
<mo>+</mo>
<msub>
<mi>&beta;</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>21</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>&upsi;</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<msub>
<mi>W</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<msubsup>
<mi>u</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
<mi>m</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>&upsi;</mi>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
<mo>+</mo>
<msub>
<mi>&alpha;</mi>
<mi>j</mi>
</msub>
<mi>&upsi;</mi>
<mo>(</mo>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>)</mo>
<mo>+</mo>
<msub>
<mi>&beta;</mi>
<mi>j</mi>
</msub>
<mi>&upsi;</mi>
<mo>(</mo>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>&OverBar;</mo>
</mover>
<mo>)</mo>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<msub>
<mi>W</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<msubsup>
<mi>u</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
<mi>m</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>+</mo>
<msub>
<mi>&alpha;</mi>
<mi>j</mi>
</msub>
<mo>+</mo>
<msub>
<mi>&beta;</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>22</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>&pi;</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<msub>
<mi>W</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<msubsup>
<mi>u</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
<mi>m</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>&pi;</mi>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
<mo>+</mo>
<msub>
<mi>&alpha;</mi>
<mi>j</mi>
</msub>
<mi>&pi;</mi>
<mo>(</mo>
<msub>
<mi>x</mi>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
</msub>
<mo>)</mo>
<mo>+</mo>
<msub>
<mi>&beta;</mi>
<mi>j</mi>
</msub>
<mi>&pi;</mi>
<mo>(</mo>
<mover>
<msub>
<mi>x</mi>
<mrow>
<mi>j</mi>
<mi>m</mi>
</mrow>
</msub>
<mo>-</mo>
</mover>
<mo>)</mo>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<msub>
<mi>W</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<msubsup>
<mi>u</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
<mi>m</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>+</mo>
<msub>
<mi>&alpha;</mi>
<mi>j</mi>
</msub>
<mo>+</mo>
<msub>
<mi>&beta;</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>23</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein vkRepresent the cluster centre of kth class;
Step 5:The image segmentation result that step 4 obtains be bianry image, prospect 1, background 0;But in obtained prospect
Nontarget area may be included;In order to extract infrared target, it is necessary to locate to the image segmentation result that step 4 obtains into follow-up
Reason rejects nontarget area;Subsequent processing is included according to herein below:
(1) region areas are less than some threshold value, then reject the region;
(2) rejects nontarget area according to the length-width ratio in region;
(3) region that rejectings are connected with image boundary.
2. it is according to claim 1 a kind of using the method for improving Intuitionistic Fuzzy Clustering algorithm extraction infrared target, it is special
Sign is:Step 1 specifically comprises the following steps:
(1) is generated infrared using Hou Xiaodi and Jonathan Halley that et al. the conspicuousness method based on image signatures proposed
The notable figure of image, gradient operator used is Sobel operators during this method extraction notable figure;
(2) splits notable figure using Otsu threshold method, and segmentation result is bianry image;Non-zero region is regarded as in segmentation result
Target region.
3. it is according to claim 1 a kind of using the method for improving Intuitionistic Fuzzy Clustering algorithm extraction infrared target, it is special
Sign is:The target is infrared pedestrian, infrared ship, infrared aircraft;According to the shape of three kinds of targets, using elliptical form
The shape of target is fitted so that the pixel within ellipse has larger γj, and the picture outside ellipse
Vegetarian refreshments has smaller γj, and distance objective center is nearer, γjIt is bigger;Symmetry axis position is determined as target's center
The row coordinate q of ordinate, i.e. elliptical center0。
4. it is according to claim 3 a kind of using the method for improving Intuitionistic Fuzzy Clustering algorithm extraction infrared target, it is special
Sign is:The aspect ratio range of infrared pedestrian is set to 1~4, infrared ship length-width ratio is set to 1~10, infrared aircraft length-width ratio
It is set to 1~3.
5. a kind of method for extracting infrared target using Intuitionistic Fuzzy Clustering algorithm is improved according to claim 1 or 3,
It is characterized in that:In infrared image, target should be highlight regions, thus the grey scale change trend of symmetry axis column for " trough-
Wave crest-trough " takes wave crest length as elliptical long axis a;Similarly, for every a line of target area, grey scale change trend
Also it is " trough-wave crest-trough ", maximum wave crest length in all rows is taken to be used as elliptical short axle b, the row coordinate in ellipse
The row coordinate p of the heart0。
6. it is according to claim 1 a kind of using the method for improving Intuitionistic Fuzzy Clustering algorithm extraction infrared target, it is special
Sign is:Used in step 4 improved Intuitionistic Fuzzy Clustering algorithm to infrared image carry out cluster segmentation the step of for:
(1) sets classification number c, maximum iteration T, iteration stopping threshold epsilon;
(2) each element u in random initializtions subordinated-degree matrix UijWith each cluster centre viDegree of membership μ (vi), non-person in servitude
Category degree υ (vi), hesitation degree π (vi);
(3) calculates each element W in weight coefficient matrix W according to formula (16), (17), (18), (19)ij;
(4) updates each element u in subordinated-degree matrix U according to formula (20)ij;
(5) calculates cluster centre v respectively according to formula (21), (22), (23)iDegree of membership μ (vi), non-affiliated degree υ (vi), still
Henan degree π (vi);
(6) if | U (t+1)-U (t) | < ε or iterations stop iteration, carry out step (6) more than maximum iteration T;
Otherwise, step (2) is carried out;T is iterations;
(7) defuzzifications complete image segmentation;
Wherein, c is the classification number in fuzzy clustering, need to manually be set;Maximum iteration T and iteration stopping threshold epsilon are used to control
Iterative process processed, when iterations be more than the T or front and rear degree of membership differences that iteration obtains twice be less than iteration stopping threshold epsilon, then
Stop iterative process.
7. it is according to claim 5 a kind of using the method for improving Intuitionistic Fuzzy Clustering algorithm extraction infrared target, it is special
Sign is:Subordinated-degree matrix U=(uij)m×n, it is one and an equal amount of matrix of image, m, n are the length and width of image, in U
Element be degree of membership uij;Random initializtion subordinated-degree matrix refers to each element u in subordinated-degree matrix UijIt is random to assign
Value, while make u after assignmentijMeet following condition:
<mrow>
<mn>0</mn>
<mo><</mo>
<msub>
<mi>u</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo><</mo>
<mn>1</mn>
<mo>,</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>c</mi>
</munderover>
<msub>
<mi>u</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>=</mo>
<mn>1</mn>
</mrow>
Similarly, weight coefficient matrix W=(Wij)m×n, it is one and an equal amount of matrix of image, the element in W is weight coefficient
Wij;Update weight coefficient matrix and subordinated-degree matrix refer to according to WijAnd uijCalculation formula calculate new WijAnd uij;
Defuzzification refers to determine each pixel x according to the subordinated-degree matrix U that step (1)-(6) obtainjGeneric
Process, pixel xjClassification determine according to the following formula:
<mrow>
<mi>L</mi>
<mi>a</mi>
<mi>b</mi>
<mi>e</mi>
<mi>l</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munder>
<mrow>
<mi>arg</mi>
<mi>max</mi>
</mrow>
<mi>i</mi>
</munder>
<mrow>
<mo>(</mo>
<msub>
<mi>u</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>24</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein, i is classification sequence number, for integer, i=1,2 ... c;Then pixel xjClassification be uijIt obtains corresponding during maximum
Classification sequence number;Subordinated-degree matrix according to obtaining determines that the process of each pixel classification is defuzzification process.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810010258.8A CN108062757B (en) | 2018-01-05 | 2018-01-05 | Method for extracting infrared target by using improved intuitionistic fuzzy clustering algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810010258.8A CN108062757B (en) | 2018-01-05 | 2018-01-05 | Method for extracting infrared target by using improved intuitionistic fuzzy clustering algorithm |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108062757A true CN108062757A (en) | 2018-05-22 |
CN108062757B CN108062757B (en) | 2021-04-30 |
Family
ID=62141272
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810010258.8A Active CN108062757B (en) | 2018-01-05 | 2018-01-05 | Method for extracting infrared target by using improved intuitionistic fuzzy clustering algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108062757B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108717069A (en) * | 2018-05-29 | 2018-10-30 | 电子科技大学 | A kind of high-pressure bottle thermal imaging imperfection detection method based on the segmentation of row variable step |
CN110211126A (en) * | 2019-06-12 | 2019-09-06 | 西安邮电大学 | Image partition method based on intuitionistic fuzzy C mean cluster |
CN110853054A (en) * | 2019-10-29 | 2020-02-28 | 北京航空航天大学 | Method for improving intuitive fuzzy clustering and extracting infrared ship by using position information |
CN111402272A (en) * | 2020-03-19 | 2020-07-10 | 西安石油大学 | Neighborhood weighted fuzzy clustering image segmentation method |
CN113011389A (en) * | 2021-04-23 | 2021-06-22 | 电子科技大学 | Road pedestrian small target detection method based on clustering idea |
CN113409335A (en) * | 2021-06-22 | 2021-09-17 | 西安邮电大学 | Image segmentation method based on strong and weak joint semi-supervised intuitive fuzzy clustering |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103020948A (en) * | 2011-09-28 | 2013-04-03 | 中国航天科工集团第二研究院二○七所 | Night image characteristic extraction method in intelligent vehicle-mounted anti-collision pre-warning system |
US8488863B2 (en) * | 2008-11-06 | 2013-07-16 | Los Alamos National Security, Llc | Combinational pixel-by-pixel and object-level classifying, segmenting, and agglomerating in performing quantitative image analysis that distinguishes between healthy non-cancerous and cancerous cell nuclei and delineates nuclear, cytoplasm, and stromal material objects from stained biological tissue materials |
CN106056581A (en) * | 2016-05-23 | 2016-10-26 | 北京航空航天大学 | Method of extracting infrared pedestrian object by utilizing improved fuzzy clustering algorithm |
CN106408569A (en) * | 2016-08-29 | 2017-02-15 | 北京航空航天大学 | Brain MRI (magnetic resonance image) segmentation method based on improved fuzzy C-means clustering algorithm |
CN106780485A (en) * | 2017-01-12 | 2017-05-31 | 西安电子科技大学 | SAR image change detection based on super-pixel segmentation and feature learning |
CN107240096A (en) * | 2017-06-01 | 2017-10-10 | 陕西学前师范学院 | A kind of infrared and visual image fusion quality evaluating method |
CN107240119A (en) * | 2017-04-19 | 2017-10-10 | 北京航空航天大学 | Utilize the method for improving the fuzzy clustering algorithm extraction uneven infrared pedestrian of gray scale |
CN107292882A (en) * | 2017-08-02 | 2017-10-24 | 国网电力科学研究院武汉南瑞有限责任公司 | One kind is based on the adaptive electrical equipment malfunction detection methods of Meanshift |
-
2018
- 2018-01-05 CN CN201810010258.8A patent/CN108062757B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8488863B2 (en) * | 2008-11-06 | 2013-07-16 | Los Alamos National Security, Llc | Combinational pixel-by-pixel and object-level classifying, segmenting, and agglomerating in performing quantitative image analysis that distinguishes between healthy non-cancerous and cancerous cell nuclei and delineates nuclear, cytoplasm, and stromal material objects from stained biological tissue materials |
CN103020948A (en) * | 2011-09-28 | 2013-04-03 | 中国航天科工集团第二研究院二○七所 | Night image characteristic extraction method in intelligent vehicle-mounted anti-collision pre-warning system |
CN106056581A (en) * | 2016-05-23 | 2016-10-26 | 北京航空航天大学 | Method of extracting infrared pedestrian object by utilizing improved fuzzy clustering algorithm |
CN106408569A (en) * | 2016-08-29 | 2017-02-15 | 北京航空航天大学 | Brain MRI (magnetic resonance image) segmentation method based on improved fuzzy C-means clustering algorithm |
CN106780485A (en) * | 2017-01-12 | 2017-05-31 | 西安电子科技大学 | SAR image change detection based on super-pixel segmentation and feature learning |
CN107240119A (en) * | 2017-04-19 | 2017-10-10 | 北京航空航天大学 | Utilize the method for improving the fuzzy clustering algorithm extraction uneven infrared pedestrian of gray scale |
CN107240096A (en) * | 2017-06-01 | 2017-10-10 | 陕西学前师范学院 | A kind of infrared and visual image fusion quality evaluating method |
CN107292882A (en) * | 2017-08-02 | 2017-10-24 | 国网电力科学研究院武汉南瑞有限责任公司 | One kind is based on the adaptive electrical equipment malfunction detection methods of Meanshift |
Non-Patent Citations (4)
Title |
---|
JIN WU ET AL: ""Infrared Image Segmentation via Fast Fuzzy C-Means with Spatial Information"", 《INTERNATIONAL CONFERENCE ON ROBOTICS AND BIOMIMETICS》 * |
XIANGZHI BAI ET AL: ""Symmetry Information Based Fuzzy Clustering for Infrared Pedestrian Segmentation"", 《IEEE TRANSACTIONS ON FUZZY SYSTEMS》 * |
刘艳丽: ""基于模糊聚类的红外图像目标分割方法研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
唐庆菊等: ""基于模糊C均值聚类和Canny算子的红外图像边缘识别与缺陷定量检测"", 《红外与激光工程》 * |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108717069A (en) * | 2018-05-29 | 2018-10-30 | 电子科技大学 | A kind of high-pressure bottle thermal imaging imperfection detection method based on the segmentation of row variable step |
CN108717069B (en) * | 2018-05-29 | 2020-08-11 | 电子科技大学 | High-pressure container thermal imaging defect detection method based on line variable step length segmentation |
CN110211126A (en) * | 2019-06-12 | 2019-09-06 | 西安邮电大学 | Image partition method based on intuitionistic fuzzy C mean cluster |
CN110211126B (en) * | 2019-06-12 | 2022-06-03 | 西安邮电大学 | Image segmentation method based on intuitive fuzzy C-means clustering |
CN110853054A (en) * | 2019-10-29 | 2020-02-28 | 北京航空航天大学 | Method for improving intuitive fuzzy clustering and extracting infrared ship by using position information |
CN110853054B (en) * | 2019-10-29 | 2022-08-16 | 北京航空航天大学 | Method for improving intuitive fuzzy clustering and extracting infrared ship by using position information |
CN111402272A (en) * | 2020-03-19 | 2020-07-10 | 西安石油大学 | Neighborhood weighted fuzzy clustering image segmentation method |
CN111402272B (en) * | 2020-03-19 | 2023-09-29 | 西安石油大学 | Neighborhood weighted fuzzy clustering image segmentation method |
CN113011389A (en) * | 2021-04-23 | 2021-06-22 | 电子科技大学 | Road pedestrian small target detection method based on clustering idea |
CN113011389B (en) * | 2021-04-23 | 2022-07-26 | 电子科技大学 | Road pedestrian small target detection method based on clustering idea |
CN113409335A (en) * | 2021-06-22 | 2021-09-17 | 西安邮电大学 | Image segmentation method based on strong and weak joint semi-supervised intuitive fuzzy clustering |
CN113409335B (en) * | 2021-06-22 | 2023-04-07 | 西安邮电大学 | Image segmentation method based on strong and weak joint semi-supervised intuitive fuzzy clustering |
Also Published As
Publication number | Publication date |
---|---|
CN108062757B (en) | 2021-04-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108062757A (en) | It is a kind of to utilize the method for improving Intuitionistic Fuzzy Clustering algorithm extraction infrared target | |
Yao et al. | An improved K-means clustering algorithm for fish image segmentation | |
Dubal et al. | Skin cancer detection and classification | |
CN105740945B (en) | A kind of people counting method based on video analysis | |
Mitianoudis et al. | Document image binarization using local features and Gaussian mixture modeling | |
CN108198193A (en) | It is a kind of to utilize the method for improving Intuitionistic Fuzzy Clustering algorithm segmentation infrared ship image | |
Feng et al. | Multiphase SAR image segmentation with $ G^{0} $-statistical-model-based active contours | |
CN107240119B (en) | The method for extracting the uneven infrared pedestrian of gray scale using fuzzy clustering algorithm is improved | |
Zhao et al. | Two-stage segmentation of unconstrained handwritten Chinese characters | |
Ji et al. | Fingerprint orientation field estimation using ridge projection | |
CN106529504B (en) | A kind of bimodal video feeling recognition methods of compound space-time characteristic | |
CN104616308A (en) | Multiscale level set image segmenting method based on kernel fuzzy clustering | |
CN110211126A (en) | Image partition method based on intuitionistic fuzzy C mean cluster | |
CN103198333A (en) | Automatic semantic labeling method of high resolution remote sensing image | |
CN103150731B (en) | A kind of fuzzy clustering image partition method | |
CN108664969B (en) | Road sign recognition method based on conditional random field | |
CN106845457A (en) | Method for detecting infrared puniness target based on spectrum residual error with fuzzy clustering | |
Beevi et al. | A robust segmentation approach for noisy medical images using fuzzy clustering with spatial probability | |
CN107564008B (en) | Rapid SAR image segmentation method based on key pixel fuzzy clustering | |
Liu et al. | Infrared ship target segmentation through integration of multiple feature maps | |
CN110853054B (en) | Method for improving intuitive fuzzy clustering and extracting infrared ship by using position information | |
Zhou et al. | Pulse-coupled neural network and its optimization for segmentation of electrical faults with infrared thermography | |
Kumar et al. | Detection of stages of melanoma using deep learning | |
CN106846321A (en) | A kind of image partition method based on Bayesian probability Yu neutral net | |
Ma et al. | A novel biologically-inspired target detection method based on saliency analysis for synthetic aperture radar (SAR) imagery |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |